These minutes have been approved by the Working Group and are now protected from editing. (See IRC log of approval discussion.)

See also: IRC log

Ian Horrocks, Boris Motik, Martin Dzbor, Peter Patel-Schneider, Bernardo Cuenca Grau, Carsten Lutz, Alan Ruttenberg, Uli Sattler, Jeremy Carroll, Ivan Herman, Ratnesh Sahay, Achille Fokoue, Michael Smith, Zhe Wu, Markus Krötzsch, Sandro Hawke, Diego Calvanese, Alan Ruttenberg
Elisa Kendall (on travel), Evan Wallace (attending Interop. Week), Rinke Hoekstra (Queen's Day), Bijan Parsia (sick), Jie Bao (conflicts to other meetings), Michael Schneider (vacation)
Ian Horrocks
Carsten Lutz


Ian Horrocks: roll call

Ian Horrocks: f2f3, my understanding is we agreed 28./29. July at MIT

Ian Horrocks: previous minutes

Peter Patel-Schneider: minutes are OK

PROPOSED: to accept previous minutes

Michael Smith: +1 to previous minutes
Zhe Wu: +1
Jeff Pan: 0 (didn't attend)

RESOLVED: accept previous minutes

Ian Horrocks: Strategy for putting up issues too complicated. New strategy: chairs look at them and decide whether they will be accepted

Pending Review Actions

Ian Horrocks: Action 140 and Action 141 (pending review) done

Overdue Actions

Ian Horrocks: Discussion of due Action 42, Action 129, Action 132 deferred since Bijan is not here

Ian Horrocks: Jeremy proposed to drop Action 135 (which was on him)

Ian Horrocks: Action 135 closed, Jeremy can create a new one later if necessary

Ian Horrocks: Action 43: Bijan suggested he might be able to help; come back next week

Alan Ruttenberg: I have pushed Bijan's actions to May 5 (so they will show as overdue when agenda planning next week)

Jeremy Carroll: I would appreciate early feedback from others in the group on preliminary review that I will provide soon


Issue 82

Ian Horrocks: Issue 82 is really two issues; one is buggy diagrams which has been fixed by Boris and Conrad

Ian Horrocks: The second part is the general issue of what is going to happen w.r.t. OMG metamodel

Ian Horrocks: I propose to close issue as first part is solved and more concrete issues can be raised later when needed

Boris Motik: Conrad was quite happy about the new diagrams.

Ian Horrocks: we talked to Evan and Elisa and Conrad and they were happy with this

PROPOSED: resolve Issue 82 as per the email

Carsten Lutz: +1
Zhe Wu: +1

RESOLVED: Issue 82

Issue 107

Ian Horrocks: Regarding Issue 107, there is a proposal from Alan not to deprecate OWL Lite because there is no real need to do that

Alan Ruttenberg: Jeremy's advice should be somewhere in the documents

Jeremy Carroll: If we believe that people should think in terms of the new profiles and not of OWL lite, we should deprecate

Jeremy Carroll: Jeremy was speaking personally - not for HP
Alan Ruttenberg: but you might be doing it, even you do deprecate it, as it is valid owl 2

Uli Sattler: we should be careful not to make OWL lite users think they have to change sth

Sandro Hawke: +1 Uli ("deprecated" may make Lite users think they have to change something)

Jeremy Carroll: agree with Uli in that the deprecation advice to OWL Lite users is "do nothing"

Uli Sattler: ...but saying that, if they want to keep working with their ontologies, they should look into the profiles would be useful as well

Ivan Herman: in other areas, deprecation means that, although it is still valid, in a later version we may make in invalid. Didn't find examples in W3C docs. But it has this aspect. Thus I am in favour of what Alan says

Jeff Pan: [Does this mean OWL Lite will be one of the OWL 2 profiles?]

Jeremy Carroll: I am convinced by Ivans arguments

PROPOSED: resolve Issue 107 as per

Carsten Lutz: +1
Alan Ruttenberg: Action to add jeremy's comments somewhere?
Zhe Wu: +1

RESOLVED: Issue 107

Alan Ruttenberg: I will put Jeremy's comments as a note in the profiles document

ACTION: Alan to put editorial note in profiles document re: Jeremy's advise about OWL Lite

Issue 47

Ian Horrocks: Issue 47 is moot since compound keys are covered by easykeys

Alan Ruttenberg: not moot - resolved

PROPOSED: Issue 47 resolved as addressed by EasyKeys

Zhe Wu: +1
Carsten Lutz: +1

RESOLVED: Issue 47

Issue 122

Ian Horrocks: Issue 122 is already resolved; wasn't resolved when raised; but now resolved by the updates of the syntax and mapping documents after the F2F

Jeremy Carroll: I have heard that people don't like it, but not from HP so no problem here

PROPOSED: resolve Issue 122 as per

Carsten Lutz: +1
Zhe Wu: +1

RESOLVED: Issue 122

Issue Discussions

Ian Horrocks: Issue Discussions

Alan Ruttenberg: also, followup on feedback re: easy keys and top/bottom from implementors

Issue 71

Ian Horrocks: we reorder issues, start with Issue 71

Jeremy Carroll: the paper mentioned in the issue was maybe too complicated

Jeremy Carroll: there are linguist theries encoded in a language tag system in a bizarre fashion (?)

Jeremy Carroll: We need access to the language tag in the model theory, which is possible neither in OWL DL nor OWL Full

Jeremy Carroll: suggestion is to provide a property in OWL with a model-theoretic treatement that allows access to tags

Jeremy Carroll: different approach is to provide a way of creating data ranges of those plain literals that have language tags that are of a particular form: a rdf:type owl:PlainLiterals; a owl:withLanguageTag "en-US"^^xsd:language .

Ian Horrocks: a subclass of literals, like englishliterals?

Boris Motik: Couldn't we use facets?
Alan Ruttenberg: question, re: Peter's mail. Can we make have a range which is plain literals without a language tag?
Boris Motik: I think we can.

Jeremy Carroll: yes; maybe we can do this using facets

Peter Patel-Schneider: plain literals without language tag is xsd:string
Alan Ruttenberg: ok, peter - good point. Need a test case for that.

Boris Motik: Interesting issue we may want to address. Two questions. First: for a bunch of objects, you have values connected via properties. If you want to single out all english literals, this can be done with facets

Boris Motik: That would allow to query all values in english

Boris Motik: Second question: do you really need to break down and get the actual language tag as a string?

Boris Motik: The latter would require n-ary datatypes

Jeremy Carroll: Refers to what is needed in the paper cited in issue description. Unions play an imporant way, e.g. for different versions of chinese

Peter Patel-Schneider: Patterns should work fine for language tags, XSD patterns

Jeremy Carroll: Seems plausible

Ian Horrocks: why are literals not always a pair of language tag and content?

Jeremy Carroll: Describes decision made by RDF group, that plainliterals without lang tag are identical to xsd:string

Alan Ruttenberg: But would like example of pattern - is this clearly documented?

Ian Horrocks: This could be the basis for a solution that Jeremy could write up

Alan Ruttenberg: I like the idea

Boris Motik: I am not sure I understood both proposals.

Jeremy Carroll: I can write up sth that is likely to be consensus

Jeremy Carroll: Will write a wiki page with texts that could then go into the documents

Ian Horrocks: Can someone do the action?

Ivan Herman: It may be worth for Jeremy to talk to Felix Sasaki

ACTION: Jeremy, Start WIki page addressing lang tag issue, liasing with I18N activity (felix?)

Issue 97

Alan Ruttenberg: we were waiting for Ivan and Sandro to explain ramifications

Ian Horrocks: suggestion from Bijan was that we only need to provide some kind of hook

Jeremy Carroll: suggest ask GRDDL WG?

Alan Ruttenberg: We need an expert to help with that

Ivan Herman: my XSLT knowledge is not good enough

Ian Horrocks: the belief is that we really need to have it

Jeremy Carroll: we can send a message to the GRDDL WG and ask for volunteers

Ivan Herman: the GRDDL WG is hibernating

ACTION: Jeremy to send mail to GRDDL groups asking for volunteer (just in case)

Ian Horrocks: bijan is not here to argue that we do not need it; in the meantime, we can ping people to help us

Ivan Herman: according to GRDDL, the transformation is not necessarily in XSLT; but all known implementations are XSLT

Alan Ruttenberg: you can be conforming in that you return a generated XSLT; but we don't need a server that does the computation

Sandro Hawke: IMHO, this is cheating

Alan Ruttenberg: the server would be hosted at W3C

Sandro Hawke: bad if you are behind firewall

Ivan Herman: a GRDDL implementation might cache the transformation

Alan Ruttenberg: good point. I concur that it doesn't work to use the trick.

Jeremy Carroll: I may not want to send my data over the internet to W3C

Jeremy Carroll: Also security issue with W3C server

Issue 71

Alan Ruttenberg: I believe the consensus was that we didn't want to handle lang in XML literals
Alan Ruttenberg: not well defined

Jeremy Carroll: XHTML would be embedded as HTML, the lang proposal does not address XML Literals, in particular XHTML embedded in OWL

Ivan Herman: this is more relevant for RDF core than for OWL WG

Feedback on EasyKeys & Top/Bottom Role

Alan Ruttenberg: zhe, jeremy and achille; feedback on easykeys and top/bottomrole

Achille Fokoue: easykeys is fine with us; top/bottomproperty we don't really see strong concrete usecases; Markus sent feedback; I still have mixed feelings

Ian Horrocks: if you have implemented OWL DL, you have implemented top/bottomrole anyway in a certain sense

Ian Horrocks: so there is no implementation as such

Alan Ruttenberg: e.g I can see forward chaining reasoners having this cost

Achille Fokoue: the question is whether we should encourage people to use that, and whether this will have performance impacts

Alan Ruttenberg: curious as well as to whether it is a performance issue

Uli Sattler: it's just syntactic sugar, no blowup

Alan Ruttenberg: ok thanks

Uli Sattler: in practical cases, things may become a bit more tricky

Jeremy Carroll: our experience is that forward chaining itself is inadequate, and the right mixture of forward- and backward-chaining is the solution

Zhe Wu: no doubt that easykeys is practical; top/bottomrole people can represent it anyway so I have no strong feeling as long as we do not put a rule in OWL-R

Achille Fokoue: question is: potential impact on profiles

Ian Horrocks: it could just not be in the profile

Alan Ruttenberg: bottom has some implication in classification, but gut is that it isn't onerous

Uli Sattler: the main motivation for having top/bottomrole was not urgent use cases; it was more to have a symmetric treatment of classes and properties

Uli Sattler: I do not think we should encourage people to use it a lot

Ian Horrocks: we should give people a bit more time to think about it

Jeremy Carroll: I am happy to resolve, with Zhe's rider not in OWL-R

Alan Ruttenberg: I would say sth stronger; there are no objections, except some fear of performance impact, but it seems we are close to consensus to add those to OWL-DL

STRAW POLL: add top/bottomproperty to OWL-DL, take actions to check profiles

Jeremy Carroll: +1 (as long as not in OWL-R)
Markus Krötzsch: +1 to add top/bottomProperty to OWL DL now
Carsten Lutz: +1
Zhe Wu: +1
Alan Ruttenberg: Add EasyKeys too, Carsten
Peter Patel-Schneider: +1 to add to DL, and let the profiles fall where they may
Diego Calvanese: +1

General Discussion: n-ary Datatypes


Uli Sattler: explains current proposal

Alan Ruttenberg: bunch of the use cases are constraints

Carsten Lutz: Two confusing points in the proposal. 1. Value space is algebraic reals, should be reals. 2. In owl 1.0 strings are required, but the proposal does not make clear whether they would be required in OWL 2, whether the binary comparisons on them are required, and how exactly they would be interpreted.

Uli Sattler: have to recheck algebraic reals

Uli Sattler: as to facets, we just took the most obvious ones

Alan Ruttenberg: Asks for decidability when we put all things together

Uli Sattler: didn't we agree that this happens everywhere in OWL and we can live with it?

Alan Ruttenberg: this brings us out of OWL DL

Markus Krötzsch: Re greater than for strings: strings in XSD are indeed unordered as per spec, though this "not preclude other applications from treating strings as being ordered"

Alan Ruttenberg: is the idea to have non-structural restrictions?

Carsten Lutz: As long as only the "required" datatypes and facets are used, no problem. Otherwise, the user is on his own. That's the only option.

Alan Ruttenberg: so we just say that: if you use optional and combine them, you are not guaranteed to be decidable

Ivan Herman: I am lost on how to express all that in RDF

Uli Sattler: one of the various issues we have to address if people like the proposal in general

Jeremy Carroll: how does the proposal constrain the interaction between these predicates and the XSD types

Uli Sattler: it doesn't constrain; as long as you use the required ones, you will be fine. When you use user-defined ones, it's your own responsibility

Jeremy Carroll: Is this proposal that we should encourage ontology designers to use mathematically more clean datatypes like real and integer, but not machine datatypes like byte and float?

Uli Sattler: yes

Michael Smith: bijan left RDF mapping open until there is support for the proposal by the group

Michael Smith: ivan, I don't really understand what the RDF rqmts might be. if you could send me an email outlining them, it'd help.

PROPOSED: carry on with n-ary datatypes?

Peter Patel-Schneider: +1 (half a datatype, half a datatype, half a datatype onward)
Carsten Lutz: +1
Zhe Wu: 0 (need more time to understand)
Diego Calvanese: +1

Jeremy Carroll: HP was very opposed; I can re-consult with colleagues to check; looks not that much better now

ACTION: Jeremy clarify HP position on n-ary datatypes

PROPOSED: should n-ary datatypes include optional datatype + predicates?

Carsten Lutz: +1
trackbot-ng: Created Action 146 - Clarify HP position on n-ary datatypes [on Jeremy Carroll - due 2008-05-07].
Diego Calvanese: +1
Peter Patel-Schneider: +1 (minus some very small epsilon)

Alan Ruttenberg: a crucial point seems to be the RDF mapping

Last modified on 7 May 2008, at 16:55